Background
Methods
Literature search
Selection criteria
Outcome assessment
Definitions and terminology
Data extraction
Statistical analysis
Quality assessment
Results
Reference (first author, year of publication) | Model | Study population | Predicted outcome | No. of participants (no. with predicted outcome) | Study interval | Source of data | Country of inclusion (no. of centres) | Model development | No. of variables screened (no. In final model) |
---|---|---|---|---|---|---|---|---|---|
Model development | |||||||||
Swindell et al. [18] | ABCD-score | Attempted CPR for IHCA in patients > 50 years | Survival to discharge | 92.706 (n.r.) | 2012–2016 | NIS Registry | USA (n.r.) | Selection of variables based on univariate analysis | n.r. (4) |
Chan et al. [19] | Benchmark hospital rates for ROSC after IHCA | Attempted CPR for IHCA | Sustained ROSC > 20 min | 83.206 (D: 55.601 (40.038), V:27.128 (19.716)) | 2014–2017 | GWTG Registry | USA (335) | Hierarchical logistic regression | 24 (10) |
Harrison et al. [20] | 1. Predicting ROSC following IHCA | Attempted CPR for IHCA | ROSC > 20 min | 22.479 (D: 14.688 (6605), V: 7791 (2926)) | 2011–2013 | NCAA Registry | UK (143) | Multilevel logistic regression | 7 (7) |
2. Predicting survival following IHCA | Attempted CPR for IHCA | Survival to discharge | 7 (6) | ||||||
Ebell et al. [22]# | 1. Classification and Regression Trees (CART1) | Attempted CPR for IHCA | Survival to discharge with CPC = 1 | 52.527 D: 38.092 (3610) V: 14.435 (1720) | 2007–2009 | GWTG Registry | USA (366) | Classification and regression trees | 32 (8) |
2. Classification and Regression Trees (CART2) | 32 (9) | ||||||||
Ebell et al. [21] | GO-FAR | Attempted CPR for IHCA | Survival to discharge with CPC = 1 | 51.240: training (44%), testing (22%), validation (34%) (overall 10.4% with outcome) | 2007–2009 | GWTG Registry | USA (366) | Previous meta-analysis; logistic regression | 17 (13) |
Chan et al. [23] | Risk-standardized survival after IHCA | Attempted CPR for IHCA | Survival to hospital discharge; risk-standardized survival rates | 48.841 (10.290) (D: 32.560 (6844), V:16.281 (3446)) | 2007–2010 | GWTG Registry | USA (356) | Hierarchical logistic regression | 26 (9) |
Larkin et al. [24] | NRCPR in-hospital mortality prediction model | Attempted CPR for IHCA | In-hospital mortality (or CPC ≥ 3) | 49,130 (84.1% with outcome) (D: 33,468 V: 12,754) | 2000–2004 | NRCPR Registry | USA (366) | Logistic regression with bootstrapping | 63 (33) |
Danciu et al. [25] | 1. Survival after IHCA | Attempted CPR for IHCA | Survival to discharge | 219 (33) | 2000–2002 | Retrospective cohort | USA (1) | Backwards stepwise logistic regression | 10 (5) |
2. Survival after IHCA | 1 month survival | 219 (30) | |||||||
3. Survival after IHCA | 3 months survival | 219 (25) | |||||||
Cooper et al. [26]** | Resuscitation Predictor Scoring Scale (RPS Scale) | Attempted CPR for IHCA | 24-h survival | 1633 (444) | 1993–2000 | Retrospective cohort | UK (1) | Multiple forward logistic regression | n.r. (3) |
Ambery et al. [27] | IHCA 5-score | Attempted CPR for IHCA | Survival to discharge | 111 (14) | 1998–1999 | Retrospective cohort | UK (1) | Previous knowledge | 5 (5) |
Dodek et al. [28] | Predictors of in-hospital mortality after IHCA | Attempted CPR for IHCA | Probability of death before hospital discharge | D: 197 (145) V:120 (90) | 1986–87; 1989–90 | Retrospective cohort | Canada (1) | Stepwise logistic regression | 31 (8) |
Ebell et al. [29] | Artificial neural networks for predicting failure to survive following IHCA | Attempted CPR for IHCA (ward + ICU) | Failure to survive | 218 (96) | 1989–1991 | Retrospective cohort | USA (1) | Artificial neural network | 24 (24) |
Lawrence et al. [30] | Probability of survival after CPR for IHCA | Attempted CPR for IHCA | Survival to discharge | 340 (88) | 1987–1989 | Retrospective cohort | USA (1) | Stepwise discriminant analysis followed by logistic regression | 18 (4) |
Marwick et al. [31]** | 1. Prognostic survival index at resuscitation attempt | Attempted CPR for IHCA (ward + CCU + ER + OR) | Survival to discharge | 710 (92) | 1984–1987 | Prospective cohort | Australia (1) | Cox multiple regression | 8 (4) |
2.Prognostic survival index after 10 min | 8 (3) | ||||||||
3.Prognostic survival index after resuscitation | 8 (5) | ||||||||
George et al. [32] | Pre-Arrest Morbidity Index (PAM) | Attempted closed-chest CPR for IHCA (ward + ICU) | ROSC, survival to discharge, 3-months survival | 140 (77, 34, 29) | 1985 | Prospective cohort | USA (1) | Previous knowledge | n.r. (15) |
Burns et al. [33] | Prediction of initial success after CPR for IHCA | Attempted CPR for IHCA (ward + ICU) | ROSC | 122 (56) | 1985 | Retrospective cohort | USA (1) | Significance testing followed by backwards logistic regression | 25 (4) |
Model updating | Model based on | |||||||||
---|---|---|---|---|---|---|---|---|---|---|
Hong et al. [34] | GO-FAR + albumin | Attempted CPR for IHCA | Survival to discharge with CPC 1 or 2 | 863 (127): D 419 (59) and V: 444 (68) | 2013–2017 | Prospective cohort | Korea (1) | Previous knowledge | 14 (14) | GO-FAR |
George et al. [35] | GO-FAR 2 | Attempted CPR for IHCA (excl. OR) | Survival to discharge with CPC ≤ 2 | 52.468: training 23.713 (44%) (4768 with outcome), testing 12.240 (22%), validation 16.515 (34%) | 2012–2017 | GWTG Registry | USA (366) | Previous knowledge; BIC and LASSO analysis | 24 (10) | GO-FAR |
Piscator et al. [36] | Prediction of outcome for In-Hospital Cardiac Arrest (PIHCA) | Attempted CPR for IHCA | Survival to discharge with CPC ≤ 2 | 717 (628 after exclusion missing cases) (174) | 2013–2014 | SCR Registry | Sweden (6) | Previous knowledge; logistic regression | 9 (9) | GO-FAR |
Dautzenberg et al. [37]* | Modified PAM index (MPI) | n.r. (according to George et al.) | n.r. (according to George et al.) | n.r. | n.r. | n.r. | n.r. | Previous knowledge | n.r. (16) | PAM |
Ebell et al. [38]* | Prognosis After Resuscitation (PAR) | n.r. (according to George et al.) | n.r. (according to George et al.) | 218 (n.r.) | n.r. | n.r. | n.r. | Previous knowledge | n.r. (8) | PAM |
Reference (first author, year of publication) | Type of factors in final model | Prognostic factors in final model | Model presentation | Evaluation of model performance | Validation | ||
---|---|---|---|---|---|---|---|
Calibration | Discrimination | Classification | |||||
Model development | |||||||
Swindell et al. [18] | Pre + intra-arrest | Age, BMI, comorbidity (cancer or liver disease), days from admission to arrest | Scoring system | n.r. | AUC 0.581 (95% CI 0.577–0.585) | ABDC ≤ 2 sensitivity 96.8% (accuracy 33.6%), ABCD ≤ -1 specificity 82% | n.r. |
Chan et al. [19] | Pre + intra-arrest | Age, initial rhythm, heart failure during admission, respiratory insufficiency, diabetes mellitus, metabolic disturbance, metastatic/haematologic malignancy, acute CNS non-stroke event, continuous IV vasopressor, mechanical ventilation | Logistic regression model with beta-coefficients estimates | D: R2 = 0.996 V: R2 = 0.990, Hosmer-Lemeshow GOF = 0.87 | D: c-statistic 0.638 V: c-statistic 0.630 | n.r. | Internal validation: split-sample |
Harrison et al. [20] | Pre + intra-arrest | 1. ROSC: age, sex, prior length of stay, reason for attendance, location of arrest, initial rhythm, interactions presenting rhythm and location | Online score calculator | Calibration plots | 1. D: c-statistic = 0.727 V: c-index 0.73. Accuracy: R2 = 0.11–0.17 | n.r. | External validation: temporal validation, geographical validation |
2. Survival: age, prior length of stay, reason for attendance, location of arrest, initial rhythm, interactions presenting rhythm and location | 2. D: c-statistic = 0.804. V: c-index 0.81. Accuracy: R2 = 0.21–0.24 | ||||||
Ebell et al. [21] | Pre-arrest | 1: CPC-score, admitting location, sepsis, mechanical ventilation, age, metastatic/haematological malignancy, acute MI this admission | Algorithm | n.r. | AUC: Model 1 D: 0.76, V: 0.73 | Classification table | External validation: temporal validation |
2: Factors of model 1 + other configuration + acute stroke | AUC: Model 2 D: 0.74, V: 0.71 | ||||||
Ebell et al. [21] | Pre-arrest | Age, admission neurologically intact, major trauma, acute stroke, metastatic/haematological cancer, septicaemia, non-cardiac admission, hypotension/hypoperfusion, respiratory insufficiency, pneumonia, metastatic cancer, renal dialysis/insufficiency, hepatic insufficiency, admitted from nursing facility | Scoring system | Hosmer–Lemeshow statistic 11.39 (p = 0.18). HL graph is shown | C-statistic overall: 0.800. Derivation and training: 0.77, validation 0.78 | Classification table | Internal validation: split-sample |
Chan et al. [23] | Pre + intra | Age, initial arrest rhythm, hospital location, hypotension, septicemia, metastatic/haematological malignancy, hepatic insufficiency, mechanical ventilation before arrest, vasopressor before arrest | Logistic regression model with beta-coefficients estimates | R2 of 0.99 in derivation and validation cohorts | C-statistic: 0.734 in derivation and 0.737 in validation cohort | n.r. | Internal validation: split-sample |
Larkin et al. [24] | Pre + intra-arrest | Code team present, age, race, illness category (medical cardiac, surgical cardiac, surgical non-cardiac and trauma, obstetrics), pre-existing conditions (MI, hypotension, hepatic insufficiency, baseline depression CNS function, acute stroke, infection/septicemia, metastatic/haematological malignancy, renal failure, major trauma), interventions in place (invasive airway, mechanical ventilation, anti-arrhythmics, vasopressors, vasodilators, chest tube), monitored outside ICU, monitored with arterial catheter, witnessed, pulseless when need for CPR recognized, event location (PACU/OR, general floor/telemetry, ED), initial pulseless rhythm (asystole vs VF, VT), admit time to event (in h) | Multivariate odds ratios | D: Hosmer-Lemeshow statistic 73.7 (p < 0.001) V: Hosmer-Lemeshow statistic 37.9 (p < 0.001) | D: AUC 0.78. V: AUC 0.77 (0.79–0.78). For data known at admission: AUC = 0.68, + pre-arrest admission known during admission AUC = 0.73 | n.r. | Internal validation: split-sample |
Danciu et al. [25] | Pre + intra-arrest | 1, 2 3: Respiratory arrest, initial rhythm, chronic renal insufficiency, higher BMI, less days from admission to resuscitation attempt | Scoring system + logistic model equation | 1. GOF p = 0.874 | n.r. | Survival to discharge: sensitivity= 0.88 specificity = 0.32 | n.r. |
2. GOF p = 0.599 | 1 month: sensitivity = 0.89 specificity = 0.31 | ||||||
3. GOF p = 0.822 | 3 months: sensitivity = 0.91 specificity = 0.32 | ||||||
Cooper et al. [26] | Pre + intra-arrest | Age, initial rhythm, primary cardiac/respiratory arrest | Scoring system | n.r. | n.r. | Accuracy 90% for 24 h survival for cases lasting > 15 min | n.r. |
Ambery et al. [27] | Pre-arrest | Cardiac history, COPD/asthma/respiratory failure, stroke, malignancy, renal insufficiency | Scoring system | n.r. | n.r. | (score of > 4) Under 75y sensitivity 83%, specificity 100%. Over 75y sensitivity 40%, specificity 85%, all sensitivity 52%, specificity 93% | n.r. |
Dodek et al. [28] | Pre + intra-arrest | Higher probability of death: Age, female gender, no. Previous arrests, electrical–mechanical dissociation. Lower: underlying coronary artery disease/valvular heart disease, VT, period July–September | Estimates from logistic model | n.r. | AUC: D 0.81, V: 0.71 | D: senitivitys + specificity: 0.75 (cut-off probability 0.75); V: sensitivity + specificity: 0.6 (cut-off probability 0.85) | External validation: temporal validation |
Ebell et al. [29] | Pre-arrest | Age, sex, heart rate, respiratory rate, FiO2, reason for admission, cancer, acute renal failure, GCS, place of residence before admission, mode of transport to hospital, white blood count, sodium, potassium, creatinine, haematocrit, temperature, MAP, pH, and others | Neural network | n.r. | AUC: 0.765 (SE = 0.048) | Sensitivity 52.1%, PPV 97% | n.r. |
Lawrence et al. [30] | Pre-arrest | Shock, abnormal BUN, abnormal PaO2, oliguria | Scoring system | n.r. | n.r. | Sensitivity 76%, specificity 65% | n.r. |
Marwick et al. [31] | Pre + intra-arrest | 1.Age, initial rhythm, CPR delay, defibrillation delay | Scoring system + regression coefficients | n.r. | AUC = 0.78 | n.r. | n.r. |
2.Age, initial rhythm, CPR delay | AUC = 0.71 | ||||||
3.Age, initial rhythm, defibrillation delay, defibrillated, intubated | AUC = 0.80 | ||||||
George et al. [32] | Pre-arrest | Hypotension, azotemia, malignancy, pneumonia, homebound lifestyle, angina pectoris, acute MI, heart failure (NYHA III or IV), S3 gallop, oliguria (< 300 ml/day), sepsis, mechanical ventilation, recent cerebrovascular event, coma, cirrhosis | Scoring system + nomogram | n.r. | n.r. | Linear correlation for ROSC: p < 0.02; survival to discharge p < 0.002; 3 month survival p < 0.002. PAM > 7: p < 0.0006 for in-hospital mortality | n.r. |
Burns et al. [33] | Pre + intra-arrest | Age, surgery scheduled before arrest, intensive care admission pre-arrest, pO2 < 8 mmHg | Scoring system and model equation | n.r. | n.r. | Sensitivity 76%, specificity 61%, accuracy 69% | n.r. |
Model updating | |||||||
Hong et al. [34] | Pre-arrest | GO-FAR + albumin | Scoring system | n.r | AUC D = 0.848 (CI = 0.802–0.893) V = 0.799 (CI = 0.745–0.853) | Net reclassification index V = 0.072 (CI 0.013–0.132) | External validation: temporal validation |
George et al. [35] | Pre-arrest | Age, admission CPC < 2, medical non-cardiac admission, surgical admission, hypotension/hypoperfusion, respiratory insufficiency, septicaemia, metastatic cancer, renal dialysis, hepatic insufficiency | Scoring system | Hosmer-Lemeshow test for calibration 21.43 (P = 0.006) | AUC training = 0.70 testing = 0.70 validation = 0.69 | Classification table | Internal validation: split-sample |
Piscator et al. [36] | Pre-arrest | Neurologically intact admission, sepsis, pneumonia, hypotension, respiratory insufficiency, medical non-cardiac admission, acute kidney injury, CCI, age | Scoring system | Calibration plot | V: AUC = 0.808 (CI 0.807–0.810) | For likelihood > 3% sensitivity = 99.4%, specificity = 8.4% | Internal validation: bootstrapping |
Dautzenberg et al. [37]* | Pre-arrest | Hypotension, uremia, malignancy, pneumonia, homebound lifestyle, angina pectoris, acute MI after 2 days, heart failure (NYHA III or IV), S3 gallop, oliguria (< 300 ml/day), sepsis, mechanical ventilation, recent cerebrovascular event, coma, age, dementia | Scoring system | n.r | n.r | n.r | n.r |
Ebell et al. [38]* | Pre-arrest | Malignancy (metastatic, non-metastatic), sepsis, dependent lifestyle, pneumonia, creatinine > 130 µmol/L, age > 70, acute MI (higher survival) | Scoring system | n.r | n.r | n.r | n.r |
Model development and updating studies
Model development | Model updating | ||||||||||||||||||||
---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
Swindell [18] | Chan [19] | Harrison [20 | Ebell [21] | Ebell [21] CART | Chan [23] | Larkin [24] | Danciu [25] | Cooper [26] | Ambery [27] | Dodek [28] | Ebell [29] | Lawrence [30] | Marwick [31] | George [32] | Burns [33] | Hong [34] | George [35] | Piscator [36] | Dautzenberg [37] | Ebell [38] | |
Pre-admission factors | |||||||||||||||||||||
Age | x | x | x | x | x | x | x | x | x | x | x | x | x | x | x | x | x | ||||
Sex | x | x | x | ||||||||||||||||||
Dependent functional status (or CPC 2) | x | x | x | x | x | x | x | x | |||||||||||||
(Metastatic/haematological) malignancy | x | x | x | x | x | x | x | x | x | x | x | x | |||||||||
Charlson Comorbidity Index | x | ||||||||||||||||||||
Chronic renal insufficiency | x | x | x | ||||||||||||||||||
Hepatic insufficiency/cirrhosis | x | x | |||||||||||||||||||
(Post-)admission factors | |||||||||||||||||||||
Respiratory insufficiency | x | x | x | x | x | x | |||||||||||||||
Renal insufficiency | x | x | x | x | x | x | x | x | x | x | |||||||||||
Hepatic insufficiency/cirrhosis | x | x | x | x | x | x | |||||||||||||||
Sepsis | x | x | x | x | x | x | x | x | x | x | |||||||||||
Heart failure/heart disease/AP | x | x | x | x | x | ||||||||||||||||
Cerebrovascular event | x | x | x | x | x | x | |||||||||||||||
Hypotension/IV vasopressor | x | x | x | x | x | x | x | x | x | x | x | ||||||||||
Mechanical ventilation | x | x | x | x | x | x | |||||||||||||||
Peri-arrest factors* | |||||||||||||||||||||
Initial rhythm | x | x | x | x | x | x | x | x | |||||||||||||
Days from admission to arrest | x | x | x | x | |||||||||||||||||
Location of arrest | x | x | x | x | x |
Model validation studies
Reference (year of publication) | Model validated | Study population (cohort type, included patients) | No. of participants (no. with predicted outcome) | Study interval | Source of data | Country of inclusion (no. of centres) | Predicted outcome | Method of validation | Evaluation of model performance | ||
---|---|---|---|---|---|---|---|---|---|---|---|
Calibration | Discrimination | Classification | |||||||||
Cho et al. [39] | GO-FAR | Attempted CPR for IHCA | 1011 (162) | 2013–2017 | Retrospective cohort | South Korea (1) | Survival to discharge with CPC 1 or 2 | Fully independent validation | n.r. | AUC: 0.81(CI 0.78–0.84) | Best at 4 for poor outcome: sensitivity 72.5% (CI 69.5–75-5), specificity 78.6% (CI 70.4–85.4), PPV 96.0% (CI 94.4–97.1), NPV 29.0% (CI 26.1–31.9) |
Thai et al. [40] | GO-FAR | Attempted CPR for IHCA | 62.131 (9347) | 2010–2016 | GWTG Registry | USA (386) | Survival to discharge with CPC = 1 | External validation; authors include initial author of score | Calibration plot (score systematically underestimates the likelihood of survival with CPC 1) | AUC: 0.75 (CI 0.748–0.758) | Classification table |
Rubins et al. [41] | GO-FAR | Attempted CPR for IHCA | 403 (133, 70) | 2009–2018 | Retrospective cohort | USA (1) | Survival to discharge and survival to discharge with CPC = 1 | Fully independent validation | n.r. | For survival to discharge with CPC = 1: AUC = 0.68(CI 0.62–0.73) | n.r. |
Piscator et al. [42] | GO-FAR | Attempted CPR for IHCA | 528 (141) (complete cases) | 2013–2014 | SCR Registry | Sweden (6) | Survival to discharge with CPC = 1 | External validation; authors include initial author of score | Calibration plot (GO-FAR model under-predicts the probability of survival with CPC = 1) | Complete cases: AUC = 0.82(CI 0.78–0.86) (AUC = 0.80 (CI 0.76–0.84) for multiple imputation analysis) | Classification table |
Limpawattama et al. [50] | PAM, PAR, MPI | Attempted CPR for IHCA | 278 (72) (166 complete IHCA cases) | 2013–2014 | Retrospective cohort | Thailand (1) | Survival to discharge | Fully independent validation | n.r. | AUC: PAM = 0.62(CI 0.51–0.73), PAR = 0.60(CI 0.49–0.71), MPI = 0.63(CI 0.52–0.73) | n.r. for only IHCA |
Guilbault et al. [43] | CART1, CART2 | Attempted CPR for IHCA | 287 (45) | 2007–2010 | Retrospective cohort | Sweden (1) | Survival to discharge with CPC = 1 | External validation; authors include initial author of score | n.r. | CART 1: AUC = 0.77. CART 2: AUC = 0.71 | Classification table |
Ohlsson et al. [44] | GO-FAR | Attempted CPR for IHCA | 287 (45) | 2007–2010 | Retrospective cohort | Sweden (1) | Survival to discharge with CPC = 1 | External validation; authors include initial author of score | n.r. | AUC: 0.85 (CI 0.78–0.91) | Best sensitivity and specificity at -3.5: 79% and 78% |
Ohlsson et al. [45] | PAM, PAR | Attempted CPR for IHCA | 287 (58) | 2007–2010 | Retrospective cohort | Sweden (1) | Survival to discharge | Fully independent validation | n.r. | AUC: PAM = 0.601(CI 0.528–0.674 p = 0.018), PAR = 0.718(CI 0.647–0.79 p = < 0.001) | Sensitivity, specificity: PAM > 5 = 21.8%, 94.8%. PAR > 5 = 31%, 94.8% |
Bowker et al. [46] | PAM, PAR, MPI | Attempted CPR for IHCA | 264 (28) | 1994–1996 | Retrospective cohort | UK (1) | Survival to discharge | Fully independent validation | n.r. | n.r | Sensitivity: PAM > 6 = 20%, PAR > 7 = 29%, MPI > 6 = 22% (specificity for all 100%) |
Ebell et al. [51] | PAM, PAR, APACHE III | Attempted CPR for IHCA | 656 (248, 35) | 2–3 years (not further specified) | Retrospective cohort | USA (3) | ROSC with transfer to ICU & survival to discharge | External validation; authors include initial author of score | n.r. | AUC: ROSC: PAR = 0.51(SE = 0.023), PAM = 0.495(SE = 0.023), APACHE III 0.495(SE = 0.023). Survival to discharge: PAR = 0.564(SE = 0.05), PAM = 0.519(SE = 0.051), APACHE III = 0.587(SE = 0.049) | n.r. |
O'keeffe et al. [48] | PAM, PAR | Attempted CPR for IHCA | 274 (25) | 2 years (not further specified) | Retrospective cohort | Ireland (1) | Survival to discharge | External validation; authors include initial author of score | n.r. | AUC: PAM = 0.67 (SE = 0.06) PAR = 0.74 (0.05) | Sensitivity: PAM > 4 = 9.2%, > 8 = 2%. PAR: > 5 = 23.7%, > 8 = 9.6%. Specificity: 100% for both scores at all cut-off points |
Cohn et al. [49] | PAM (minus S3 gallop) | Survivors of attempted CPR for IHCA and matched non-survivors | 86 (43) | 1986–1991 | Retrospective case–control | USA (1) | Survival to discharge | Fully independent validation | n.r. | p < 0.0003; R2 = 9.6% | n.r. |